Building predictive Markov models of ion channel permeation from molecular dynamics simulations.

Abstract

Molecular dynamics (MD) simulation of biological processes has always been a challenging task due to the long timescales of the processes involved and the large amount of output data to handle. Markov state models (MSMs) have been introduced as a powerful tool in this area of research, as they provide a mechanistically comprehensible synthesis of the large amount of MD data and, at the same time, can be used to rapidly estimate experimental properties of biological processes. Herein, we propose a method for building MSMs of ion channel permeation from MD trajectories, which directly evaluates the current flowing through the channel from the model's transition matrix (T), which is crucial for comparing simulations and experimental data. This is achieved by including in the model a flux matrix that summarizes information on the charge moving across the channel between each pair of states of the MSM and can be used in conjunction with T to predict the ion current. A procedure to drastically reduce the number of states in the MSM while preserving the estimated ion current is also proposed. Applying the method to the KcsA channel returned an MSM with five states with significant equilibrium occupancy, capable of accurately reproducing the single-channel ion current from microsecond MD trajectories.